Download exploring anticorrelations and light element variations

The Astronomical Journal, 149:153 (24pp), 2015 May
doi:10.1088/0004-6256/149/5/153
© 2015. The American Astronomical Society. All rights reserved.
EXPLORING ANTICORRELATIONS AND LIGHT ELEMENT VARIATIONS IN
NORTHERN GLOBULAR CLUSTERS OBSERVED BY THE APOGEE SURVEY
Szabolcs Mészáros1,2, Sarah L. Martell3, Matthew Shetrone4, Sara Lucatello5, Nicholas W. Troup6, Jo Bovy7,
Katia Cunha8,9, Domingo A. García-Hernández10,11, Jamie C. Overbeek2, Carlos Allende Prieto10,11,
Timothy C. Beers12, Peter M. Frinchaboy13, Ana E. García Pérez6,10,11, Fred R. Hearty14, Jon Holtzman15,
Steven R. Majewski6, David L. Nidever16, Ricardo P. Schiavon17, Donald P. Schneider14,18, Jennifer S. Sobeck6,
Verne V. Smith19, Olga Zamora10,11, and Gail Zasowski20
1
ELTE Gothard Astrophysical Observatory, H-9704 Szombathely, Szent Imre Herceg st. 112, Hungary
2
Department of Astronomy, Indiana University, Bloomington, IN 47405, USA
Department of Astrophysics, School of Physics, University of New South Wales, Sydney, NSW 2052, Australia
4
University of Texas at Austin, McDonald Observatory, Fort Davis, TX 79734, USA
5
INAF-Osservatorio Astronomico di Padova, vicolo dell Osservatorio 5, I-35122 Padova, Italy
6
Department of Astronomy, University of Virginia, Charlottesville, VA 22904-4325, USA
7
Institute for Advanced Study, Einstein Drive, Princeton, NJ 08540, USA
8
University of Arizona, Tucson, AZ 85719, USA
9
Observatório Nacional, São Cristóvão, Rio de Janeiro, Brazil
10
Instituto de Astrofísica de Canarias (IAC), E-38200 La Laguna, Tenerife, Spain
11
Universidad de La Laguna, Departamento de Astrofísica, E-38206 La Laguna, Tenerife, Spain
12
Department of Physics and JINA Center for the Evolution of the Elements, University of Notre Dame, Notre Dame, IN 46556, USA
13
Texas Christian University, Fort Worth, TX 76129, USA
14
Department of Astronomy and Astrophysics, The Pennsylvania State University, University Park, PA 16802, USA
15
New Mexico State University, Las Cruces, NM 88003, USA
16
Department of Astronomy, University of Michigan, Ann Arbor, MI 48109, USA
17
Astrophysics Research Institute, IC2, Liverpool Science Park, Liverpool John Moores University, 146 Brownlow Hill, Liverpool, L3 5RF, UK
18
Institute for Gravitation and the Cosmos, The Pennsylvania State University, University Park, PA 16802, USA
19
National Optical Astronomy Observatory, Tucson, AZ 85719, USA
20
Johns Hopkins University, Baltimore, MD 21218, USA
Received 2014 November 5; accepted 2015 January 20; published 2015 April 14
3
ABSTRACT
We investigate the light-element behavior of red giant stars in northern globular clusters (GCs) observed by the
SDSS-III Apache Point Observatory Galactic Evolution Experiment. We derive abundances of 9 elements (Fe, C,
N, O, Mg, Al, Si, Ca, and Ti) for 428 red giant stars in 10 GCs. The intrinsic abundance range relative to
measurement errors is examined, and the well-known C–N and Mg–Al anticorrelations are explored using an
extreme-deconvolution code for the first time in a consistent way. We find that Mg and Al drive the population
membership in most clusters, except in M107 and M71, the two most metal-rich clusters in our study, where the
grouping is most sensitive to N. We also find a diversity in the abundance distributions, with some clusters
exhibiting clear abundance bimodalities (for example M3 and M53) while others show extended distributions. The
spread of Al abundances increases significantly as cluster average metallicity decreases as previously found by
other works, which we take as evidence that low metallicity, intermediate mass AGB polluters were more common
in the more metal-poor clusters. The statistically significant correlation of [Al/Fe] with [Si/Fe] in M15 suggests that
28
Si leakage has occurred in this cluster. We also present C, N, and O abundances for stars cooler than 4500 K and
examine the behavior of A(C+N+O) in each cluster as a function of temperature and [Al/Fe]. The scatter of A(C+N
+O) is close to its estimated uncertainty in all clusters and independent of stellar temperature. A(C+N+O) exhibits
small correlations and anticorrelations with [Al/Fe] in M3 and M13, but we cannot be certain about these relations
given the size of our abundance uncertainties. Star-to-star variations of α-element (Si, Ca, Ti) abundances are
comparable to our estimated errors in all clusters.
Key words: stars: abundances – stars: AGB and post-AGB – stars: chemically peculiar – stars: evolution
Supporting material: machine-readable and VO tables
1. INTRODUCTION
recognized to be the signature of high-temperature H-burning.
This was initially framed within a stellar evolutionary scenario
(see, e.g., Kraft 1994, and references therein) given that GC
abundance work based on high-quality data was limited to
bright, evolved giants. It was only at the turn of the century that
the availability of high-resolution spectrographs mounted on
8 m class telescopes made it possible to carry out studies on the
compositions of stars down to the main-sequence, which
revealed light element variations analogous to those found
among giants (Briley et al. 1996; Gratton et al. 2001; Ramírez
& Cohen 2002). Given that the atmospheres of warm main-
Over the last two decades, the long-lasting idea of globular
clusters (GCs) hosting single, simple stellar populations has
changed dramatically. The classical paradigm of GCs being an
excellent example of a simple stellar population, defined as a
coeval and initially chemically homogeneous assembly of stars,
has been challenged by observational evidence. The presence
of chemical inhomogeneities, in most cases limited to the light
elements (the chemical pairs C-N, O-Na, and Mg-Al anticorrelated with each other), have been known for decades and
1
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
three-year, near-infrared (15090–16990 Å; Wilson et al. 2012),
high-resolution spectroscopic survey of about 100,000 red
giant stars included as part of the third Sloan Digital Sky
Survey (SDSS-III Eisenstein et al. 2011). With a nominal
resolving power of 22,500, APOGEE is deriving abundances of
up to 15 elements for nearly 100,000 stars, although fewer
elements are generally detected in weak-lined metal-poor stars.
APOGEE is in a unique position among the various Galactic
spectroscopic surveys such as Gaia-ESO, (Gilmore et al.
2012), RAVE (Steinmetz et al. 2012), and GALAH (Freeman 2012), as it uses the Sloan 2.5 m telescope at Apache Point
Observatory (Gunn et al. 2006), and thereby has access to the
northern hemisphere. APOGEE observes a large sample of
northern GCs, something that makes it possible to analyze
these clusters in a homogeneous way, which has not been done
before for these objects.
The study of GCs with APOGEE plays an important role not
just because it has access to many northern GCs. Its highresolution near-IR spectra allow the simultaneous determination of many elemental abundances generally not available in
optical spectroscopic work of GC stars. C and N, which are
elements heavily affected by the pollution phenomenon in GCs,
are often not included in studies of metal-poor stars because the
strongest features (CH and CN) lie in the near-UV, far from the
optical lines of Na, Mg, and Al, and thus multiple detectors or
setups are required to obtain both sets. In addition, because
these studies usually focus on fairly red stars, longer exposure
times are required to acquire sufficient supernova remnants to
analyze the near-UV features.
The spectra used in this paper are publicly available as part
of the 10th data release (DR10, Ahn et al. 2014) of SDSS-III.
The initial set of stars selected were the same used by Mészáros
et al. (2013) to check the accuracy and precision of APOGEE
parameters published in DR10. However, instead of using the
automatic ASPCAP pipeline, we will make use of photometry
and theoretical isochrones to constrain the effective temperature (T eff ) and surface gravity log g and use an independent
semi-automated method for elemental abundance determination
for 10 northern GCs. Some of these clusters are well studied,
such as M3, M13, M92, and M15, while others have been
poorly studied (NGC 5466), or have been only recently
discussed in the literature, such as M2 (Yong et al. 2014).
sequence stars in large part retain the composition of the gas
from which they were formed, the unavoidable conclusion was
that the abundance inhomogeneities are of primordial origin.
The most extensive spectroscopic survey of GCs undertaken
so far (Carretta et al. 2009a, 2009b, 2009c) revealed that these
inhomogeneities are ubiquitous in Galactic GCs, though they
do not appear to occur in other star formation environments.
However, the extent of the inhomogeneity varies from cluster
to cluster, and appears to correlate strongly with the presentday total mass of the GCs, and also with metallicity. The
improvement in available instrumentation and techniques has
also led to the discovery of a much higher degree of complexity
of GC color–magnitude diagrams. In fact, while some clusters
seem to photometrically comply with the simple stellar
population paradigm, a growing number of them are found to
be characterized by multiple main sequences and/or subgiant
and/or giant branches (Piotto et al. 2007; Milone et al. 2008,
e.g.,), which have been associated with variations in the
content of He and CNO, as well as age spread (DAntona
et al. 2005; Cassisi et al. 2008).
This observational evidence led to a general scenario where
GCs host multiple stellar populations. These are often assumed
to be associated with different stellar generations: the ejecta of
the slightly older stars, probably mixed with varying amounts
of gas from the original star-forming cloud, creates a
subsequent younger generation of stars (see e.g., Decressin
et al. 2007; D’Ercole et al. 2008), though alternative scenarios
are also being considered (see, e.g., Bastian et al. 2013). It is
believed that only a fraction of the first generation of stars can
contribute to the internal enrichment. The difference in ages
among the stellar generations is actually relatively small for the
majority of the clusters (with a few exceptions such as, e.g., ω
Cen or M22) and is confined to a couple of hundreds of Myr.
The details of this formation scenario are still far from being
understood. The origin of the polluting material remains to be
established and it has obvious bearings on the timescales for
the formation of the cluster itself and its mass budget. The
observed wide star-to-star variations in C, N, O, Na, and Al
found in each Galactic GC, coupled with the uniformity in Fe
and Ca (apart from a few notable exceptions), provide quite
stringent constraints and argue against anything but a minor
contribution from supernovae (Carretta et al. 2009a). Proposed
candidate polluters include intermediate mass stars in their
asymptotic giant branch (AGB) phase (Ventura et al. 2001),
fast rotating massive stars losing mass during their main
sequence phase (Decressin et al. 2007), novae (Maccarone &
Zurek 2012), and massive binaries (de Mink et al. 2009).
These potential contributions obviously operate on different
timescales and require a different amount of stellar mass in the
first generation. All of the candidates proposed so far fall short
of reproducing the full variety of observations. Advances in the
theoretical modeling of star formation and evolution are likely
needed to improve our understanding of these issues, including
the spanning of a larger range of the parameter space (e.g.,
mass, metallicity, and mass loss). However, from an observational point of view, the increase in the high-quality abundance
work available for GCs, both in the sheer number of stars and
clusters, as well as in terms of chemical species considered, is
paramount, as it creates a more complete picture of the
phenomena involved.
The Apache Point Observatory Galactic Evolution Experiment (APOGEE; S.R. Majewski et al. 2015, in preparation) is a
2. ABUNDANCE ANALYSIS
2.1. Target Selection
Table 1 lists the GCs APOGEE observed in its first year,
along with the adopted [Fe/H], E(B−V), and ages from the
literature. Targets were selected as cluster members if (1) there
is published abundance information on the star as a cluster
member, (2) the star is a radial velocity member, or (3) if it has
a probability >50% of being a cluster member based on their
proper motion adopted from the literature. After this initial
selection, we checked the position of stars in the T eff - log g
diagram based on APOGEE observations and deleted those that
were not red giant branch (RGB) stars. Stars that have
metallicity 0.3 dex (typically 3σ scatter) larger or smaller than
the cluster average also need to be deleted, but this last step
resulted in no rejections. The cluster target selection process is
described in more detail in Zasowski et al. (2013). The final
sample consists of 428 stars from 10 GCs. High signal-to-noise
ratio (S/N) spectra are essential to determine abundances from
2
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Table 1
Properties of Clusters from the Literature
systematic differences, in the range of 100–200 K, are present
between the ASPCAP and photometric temperatures. The
ASPCAP DR10 T eff were also compared to literature spectroscopic temperatures, and the average of these differences were
found to be negligible. The corrected ASPCAP DR10
temperatures were calculated between 3500 and 5500 K using
a calibration relation derived from the comparison with the
González Hernández & Bonifacio (2009) scale. ASPCAP
DR10 raw temperatures above 5000 K in metal-poor stars
showed significant, 300−500 K offsets compared to photometry, and are thus believed to be not accurate enough for
abundance analysis. This issue is mostly limited to metal-poor
stars, and does not affect stars at [Fe/H]> -1, where the vast
majority of APOGEE targets are.
The adoption of a purely photometric temperature scale
enables us to be somewhat independent of ASPCAP (while still
using the same spectra), which gives important comparison
data for future ASPCAP validation. Besides providing an
independent comparison data set for APOGEE, the photometric
temperatures allowed us to include stars in the sample that are
hotter than 5000 K. Because of these reasons, the final results
presented in this paper are based on the photometric
temperatures, and we only use the ASPCAP DR10 raw
temperatures to estimate our errors related to the atmospheric
parameters.
ID
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
7078
6341
5024
5466
6205
7089
5272
5904
6171
6838
Name
Na
[Fe/H]b
M15
M92
M53
23
47
16
8
81
18
59
122
42
12
−2.37 ± 0.02
−2.31 ± 0.05
−2.10 ± 0.09
−1.98 ± 0.09
−1.53 ± 0.04
−1.65 ± 0.07
−1.50 ± 0.05
−1.29 ± 0.02
−1.02 ± 0.02
−0.78 ± 0.02
M13
M2
M3
M5
M107
M71
E(B−V) Ref.a
0.10
0.02
0.02
0.00
0.02
0.06
0.01
0.03
0.33
0.25
a
N is the number of stars observed in each cluster.
[Fe/H] references: Harris (1996, 2010 edition) clusters are listed in order of
the average cluster metallicity determined in this paper.
c
E(B−V) references: (1) Harris (1996, 2010 edition).
b
atomic and molecular features, thus all selected targets have at
least S/N = 70 as determined by Mészáros et al. (2013).
2.2. Atmospheric Parameters
Abundances presented in this paper are defined for each
individual element X heavier than helium as
[X H] = log10 ( n X n H)star - log10( n X n H)
(1)
2.4. The Metallicity
The APOGEE DR10 release contains metallicities derived
by ASPCAP for all stars analyzed, thus providing an alternative
scale to manually derived metallicities. That metallicity, [M/H],
tracks all metals relative to the Sun, and gives the overall
metallicity of the stars because it was derived by fitting the
entire wavelength region covered by the APOGEE spectrograph. This is different from most literature publications that
use Fe lines to track metallicity in a stellar atmosphere. We use
[Fe/H] in this paper whenever we refer to metallicity presented
here, because we use Fe I lines to measure it. For the most part,
one can treat values of [M/H] as if they were [Fe/H]. When
ASPCAP metallicity was compared with individual values
from high-resolution observations from the literature, a
difference of 0.1 dex is found below [M/H] = −1, and this
discrepancy increased with decreasing metallicity reaching
0.2–0.3 dex around [M/H] = −2 (Mészáros et al. 2013). The
calibrated DR10 metallicities map well onto [Fe/H], because
the calibration process uses [Fe/H] values from the optical.
Because an alternative metallicity based on only iron lines
did not exist for APOGEE in DR10, and because of these small
zero point offsets, we decided to derive our own [Fe/H] based
on Fe I lines found in the H band, instead of using the published
APOGEE DR10 [M/H] values. We assumed that all stars have
the literature cluster mean metallicity before starting the
calculations, after which small wavelength windows around
the Fe lines listed in Table 3 were used to revise the individual
star metallicities.
where nX and nH are respectively the number of atoms of
element X and hydrogen, per unit volume in the stellar
photosphere.
To derive abundances from stellar spectra, we first have to
estimate four main atmospheric parameters: T eff , log g ,
microturbulent velocity, and overall metallicity ([Fe/H]). In
the following sub-sections we present our methodology for
determining these parameters and our reasons for not using the
values available for each star from the APOGEE Stellar
Parameters and Chemical Abundances Pipeline (ASPCAP;
García Pérez et al. 2015, in preparation) in DR10. In order to
evaluate the accuracy and precision of the ASPCAP parameters, Mészáros et al. (2013) carried out a careful comparison
with literature values using 559 stars in 20 open and GCs.
These clusters were chosen to cover most of the parameter
range of stars APOGEE is expected to observe. Mészáros et al.
(2013) provided a detailed explanation of the accuracy and
precision of these parameters, and also derived empirical
calibrations for T eff , log g, and [M/H] using literature data. In
the sections below we will briefly review these calibrations
along with their limitations.
2.3. The Effective Temperature
We adopt a photometric effective temperatures calculated
from the J−Ks colors using the equations of González
Hernández & Bonifacio (2009). Their calibration was chosen
because of its proximity of only 30−40 K to the absolute
temperature scale. De-reddened J−Ks were calculated the same
way as by Mészáros et al. (2013), from E(B−V), listed in
Table 1 for each cluster, using E (J-K s ) = 0.46 · E (B-V ).
The ASPCAP DR10 effective temperatures were compared
with photometric ones using calibrations by González Hernández & Bonifacio (2009) based on 2MASS J−Ks colors
(Strutskie et al. 2006). Mészáros et al. (2013) found that small
2.5. The Surface Gravity
In this study we adopt gravities from stellar evolution
calculations. Following Mészáros et al. (2013), we derive
gravities for our sample using isochrones from the Padova
group (Bertelli et al. 2008, 2009). The cluster metallicities
collected from the literature used in the isochrones are listed in
3
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Table 2
Properties of Stars Analyzed
2MASS ID
2M21301565+1208229
2M21301606+1213342
2M21304412+1211226
2M21290843+1209118
2M21294979+1211058
Cluster
v helio
T eff
log g
[Fe/H]
[C/Fe]
[N/Fe]
[O/Fe]
M15
M15
M15
M15
M15
−104.5
−108.3
−102.7
−106.0
−107.6
4836
4870
4715
4607
4375
1.56
1.64
1.28
1.03
0.56
−2.12
−2.31
−2.12
−2.07
−2.31
K
K
K
K
−0.44
K
K
K
K
0.95
K
K
K
K
0.44
[Mg/Fe]
[Al/Fe]
[Si/Fe]
0.16
0.10
−0.45
−0.11
0.17
−0.06
0.57
0.63
0.75
0.64
0.35
0.46
0.60
0.41
0.44
[Ca/Fe]
[Ti/Fe]
0.19
0.53
0.35
K
0.06
K
K
K
K
K
(This table is available in its entirety in machine-readable and Virtual Observatory (VO) forms.)
Table 1, while the ages of all clusters were chosen to be 10 Gyr.
The final set of temperatures and gravities corresponding to
them from the isochrones are listed in Table 2.
The ASPCAP DR10 surface gravities were compared to both
isochrones and Kepler (Borucki et al. 2010) asteroseismic
targets observed by APOGEE (Pinsonneault et al. 2014) to
estimate their accuracy and precision. An average difference of
0.3 dex was found at solar metallicity in both cases, but this
increased to almost 1 dex for very metal-poor stars. The
asteroseismic gravities are believed to have errors in the range
of 0.01−0.05 dex, thus far superior to spectroscopic measurements, but they are only available for metal-rich stars with
[M H] > -1.0 . The discrepancy between spectroscopic and
asteroseismic surface gravities is a topic of ongoing investigation (e.g., Epstein et al. 2014; Pinsonneault et al. 2014). The
final calibration from Mészáros et al. (2013) combined surface
gravities derived from isochrones below [M/H] < -1.0 with
the asteroseismic data set for high metallicities. Values after the
calibration still show some small (<0.1 dex) offsets and large
scatter around the isochrones for the GCs. Thus, in order to
minimize errors in abundances related to the uncertainty in the
ASPCAP corrected spectroscopic gravities, we decided to use
the pure isochrone gravities in this study.
DR12, APOGEE’s next public data release; Alam et al. 2015;
Holtzman et al. 2015), and includes atomic and molecular
species. The molecular line list is a compilation of literature
sources including transitions of CO, OH, CN, C2, H2, and SiH.
All molecular data are adopted without change with the
exception of a few obvious typographical corrections. The
atomic line list was compiled from a number of literature
sources and includes theoretical, astrophysical, and laboratory
oscillator strength values. These literature line positions,
oscillator strengths, and damping values were allowed to vary
in order to fit to the solar spectrum and the spectrum of
Arcturus, thus generating a tuned astrophysical line list. The
solution is weighted such that the solar solution has twice the
weight as the Arcturus solution to properly consider the fact
that the abundance ratios in Arcturus are more poorly
understood than those of the Sun. The code used for this
process was based on the LTE spectral synthesis code MOOG
(Sneden 1973) but adapted to our unique needs. For lines with
laboratory oscillator strengths, we did not allow the astrophysical gf value to vary beyond twice the error quoted by the
source. A more detailed description of this process and the line
list can be found in M. Shetrone et al. (2015, in preparation).
The choice for the local continuum set can greatly affect the
derived abundances, thus we needed a reliable automated way
to determine the continuum placement. This was done with a
separate c 2 minimization from the one that was used for the
abundance determination. Continuum normalized observation
points around 1 are multiplied by a factor between 0.7 and 1.1
with 0.001 steps for each synthesis emulating slightly different
choices for the location of the local continuum. Multiplication
is necessary because it preserves the original spectrum. The c 2
near the continuum is calculated and compared to the
continuum of the observation and minimized separately for
every abundance step. The c 2 calculation for the abundances
determination happens between certain flux ranges (usually
between 0.3 and 1.1) using the continuum placement
determined in the previous step.
2.6. The Microturbulent Velocity
A relation between microturbulent velocity and surface
gravity, vmicro = 2.24 - 0.3 ´ log g , was used for ASPCAP
analyses in DR10 (Mészáros et al. 2013). For consistency with
ASPCAP, and for easier future comparisons with APOGEE
results, we adopted that equation in this work.
3. AUTOSYNTH
The program called autosynth was developed especially for
this project to simplify the large amount of synthesis required
for abundance determination. The program takes atmospheric
parameters as input and carries out spectral synthesis to derive
elemental abundances. The core of autosynth is MOOG201321
(Sneden 1973), which does the spectrum synthesis, while
autosynth compares the synthetic spectrum with the observed
spectrum and determines the best abundances with c 2
minimization in wavelength windows specified by the user.
The program can read the MOOG formatted ATLAS and
MARCS model atmospheres and can also convert the original
ATLAS and MARCS formats into a MOOG compatible
format.
The line list adopted for this study includes both atomic and
molecular species. It is an updated version of what was used for
the DR10 results, version m201312160900 (also used for
21
3.1. Individual Abundances
The individual abundances were determined using the 1D
Local Thermodynamic Equilibrium (LTE) model atmospheres
calculated with ATLAS9 (Kurucz 1979). The model atmospheres were generated using solar reference abundances from
Asplund et al. (2005), the same way as the main APOGEE
model atmosphere database was generated (Mészáros et al.
2012). Because the overall metallicity of these clusters were
well known from the literature, initially we calculated atmospheres using the average literature metallicity for each cluster
adopting the photometric effective temperatures and isochrone
http://as.utexas.edu/ chris/moog.html
4
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
2. Individual Fe I lines are fit with autosynth, and an average
[Fe/H] is calculated for each star.
3. A new model atmosphere is calculated using this new
[Fe/H] value, but still using the starting CNO abundances.
4. We set the abundances of C, N, and O before the
remaining elements, because they can have a significant
effect on the atmospheric structure in cool stars. Since
molecular features generally disappear from metal-poor
spectra above 4500 K, we divide our stars into two
temperature groups. For the stars cooler than 4500 K, we
first determine [O/Fe] using OH lines, then create a new
model atmosphere with [α/Fe] equal to [O/Fe]. We then
determine C and O abundances from CO lines, then
recreate the model atmosphere again with these new [C/
Fe] and [O/Fe] abundances. Finally, we derive N
abundance using CN lines. For stars hotter than
4500 K, we leave the C, N, and O abundances at their
inital values.
5. The abundances of the remaining elements (Mg, Al, Si,
Ca, and Ti) are determined with autosynth, using the
stellar parameters, metallicities, and C, N, and O
abundances previously determined.
Table 3
Wavelength Regions
Element
log(N)a
Fe
7.45
C
8.39
N
O
7.78
8.66
Mg
Al
Si
7.53
6.37
7.51
Ca
Ti
6.31
4.90
a
b
Wavelength ( Å)b
15210–15213.5; 15397–15401; 15651–15654
15966–15973; 16044–16048; 16156–16160
16168–16171
15572–15606; 15772–15791; 15980–16037
16172–16248; 16617–16677; 16839–16870
15240–15417
15267–15272; 15281–15288; 15372–15380
15386–15390; 15394–15397; 15404–15414
15499–15502; 15508–15511; 15539–15542
15561–15566; 15569–15574; 15887–15904
16188–16198; 16207–16213; 16233–16237
16244–16247; 16251–16261; 16300–16305
16314–16319; 16707–16714; 16718–16720
16731–16735; 16888–16892; 16898–16912
15741–15757; 15767–15773
16720–16727; 16751–16759; 16765–16770
15962–15966; 16062–16066; 16097–16101
16218–16223; 16683–16687; 16830–16834
16139–16143; 16153.5–16164
15546.5–15549.5; 15718–15721.5
For each element, we average together the abundance results
from the different wavelength regions to obtain final values.
Although the size of each region is different, we did not find it
necessary to use weights based on their ranges or line strengths,
because that approach did not produce abundances significantly
different from a straightforward average. Data reduction errors
or missing data affected some of these regions, resulting in
erroneous fits, and because of this we carefully examined each
fit by eye. These wavelength regions were not included when
constructing the final average abundances. The final abundance
values are listed in Table 2.
The solar reference abundances are from Asplund et al. (2005).
Vacuum wavelength.
gravities. These initial model atmospheres were later revised to
have consistency with the synthesis.
The windows used to derive the individual abundances were
determined based on the analysis of FTS stars in the H-band
using the APOGEE line list by Smith et al. (2013). In the case
of Fe we measured [Fe/M], relative to the literature cluster
metallicity for each line. The abundance of Na is very
important in discussing the spread of O in GCs, and two Na
lines are available in the APOGEE spectral band. However,
these two Na lines are weak even at solar metallicities. We
carried out a number of tests attempting to derive Na
abundances, but we found that the two Na lines become very
weak around [Fe/H] = −0.5, and non-detectable below about
−0.7, thus we were not able to determine Na abundances for
any of our targets. The list of wavelength regions used in our
analysis and the solar reference values for each element are
listed in Table 3. Figures 1 and 2 show examples of observed
Fe, Mg, Al, OH, CO, and CN line profiles and their fitted
synthesis for one star from M71 and M13. The wavelength
regions shown in these figures are only a fraction of what has
been used from Table 3.
CN lines spread over most of the H band, hence it is
important to calculate the CNO abundances before the atomic
ones. It is also important to use self-consistent model
atmospheres because stars in GCs exhibit low carbon and high
α content, which significantly alters the structure of the
atmosphere compared to a solar scaled one (Mészáros et al.
2012). Taking into account all this, we developed the following
procedure to produce the final abundances for each star:
3.2. Uncertainty Calculations
3.2.1. Systematic Uncertainties
The uncertainty in the atmospheric parameters strongly
affects the final abundances derived from some of the spectral
features we consider. To test the sensitivity of abundances due
to changes in the atmospheric parameters we used the results
from the ASPCAP raw temperature scale. The same exact steps
described in the previous section were followed, but instead of
adopting the photometric temperature scale we adopt the
ASPCAP DR10 raw temperature scale, which results in new
surface gravities and microturbulent velocities. This way we
could track systematics uncertainties sensitive to these parameters as well.
The differences in abundances as a function of photometric
temperatures are demonstrated in Figure 3. The top left panel
displays the differences in the measured abundances by using
ASPCAP and photometric temperature, while the rest of the
panels are assigned to each element. The color scale in all
panels represents ΔT eff . We defined the estimated errors
associated with the atmospheric parameters based on the
standard deviation around the mean differences between the
two temperature scales. The calculated standard deviation of
the difference in temperatures is 146 K (which we round to
150 K). This standard deviation corresponds to the sum of the
uncertainty in the photometric temperature and the ASPCAP
temperature in quadrature.
1. A model atmosphere is generated using literature cluster
average metallicities, the photometric temperature, and an
isochrone gravity. Because all of our targets are RGB
stars, we choose [C/Fe] = −0.5, [O/Fe] = 0.3, and [N/
Fe] = 0.5 dex for this initial model.
5
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Figure 1. Example spectra and the fitted synthesis of Fe, Al, and Mg lines for two stars from M71 and M13. Abundances were fitted between the labeled minimum
and maximum values using a step of 0.01 dex. The printed best fitted abundance values might not be the same as in Table 2 because the table contains averaged values,
not individual fits.
In order to estimate the uncertainty of just the photometric
temperature component, we carried out calculations of
temperatures with varied J−Ks colors, reddenings, and
metallicities for M107. M107 was chosen because it is the
cluster with the highest reddening in our sample. The
uncertainty of 2MASS photometry is usually between 0.025
and 0.03 mag for these stars, and by using 0.03 as a baseline,
we estimate an uncertainty of 0.05 magnitude for the J−Ks
color. Changing J−Ks by 0.05 mag typically produces a
change of 80 K in the photometric temperature. The reddening of M107 was changed by 0.03, simulating a 10%
uncertainty in reddening, and this produced a difference of
about 40 K in the photometric temperature. A change of
0.1 dex in metallicity results in 1 K or less uncertainty in
temperature; thus the uncertainties in metallicity can be
neglected. By adding the uncertainty from photometry and
reddening in quadrature, we estimate the uncertainty of the
photometric temperature to be about 100 K. The top panel of
6
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Figure 2. Example spectra and the fitted synthesis of OH, CO, and CN lines for two stars from M71 and M13. For more explanation see caption of Figure 1 and
Section 3.1.
Table 4 lists uncertainties associated with both 150 and 100 K
changes in temperature.
In our methodology a 100 K change in temperature also
introduces an 0.3 dex systematic difference in surface gravity,
and 0.1 km s−1 in microturbulent velocity, so by coupling these
two parameters to T eff , our systematic uncertainties include the
investigation of abundance sensitivity to these parameters
as well.
3.2.2. Internal Uncertainties
Besides the uncertainty coming from the adopted atmospheric
parameters, the uncertainty of the fit was also calculated using the
σ of the residuals between the best fit synthesis and the
observation. These calculations estimate random uncertainties.
This σ of the fit is calculated within the windows used in the c 2
calculation. For determining the uncertainty of the fit, we
multiply the observed spectra by 1 + σ and 1 − σ which simulates
7
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Table 4
Estimated Abundance Uncertainties
Cluster
Fe
C
N
O
Mg
Al
Si
Ca
Ti
Systematic Uncertainties from Atmospheric Parameters
ΔT eff = 150 K
0.08
0.12
0.16
0.18
0.04
0.07
0.05
0.06
0.15
ΔT eff = 100 K
0.05
0.08
0.11
0.12
0.03
0.05
0.03
0.04
0.10
0.16
0.12
0.14
0.03
0.12
0.10
0.10
0.06
0.08
0.05
0.13
0.12
0.11
0.07
0.07
0.09
0.08
0.07
0.07
0.07
0.19
0.22
0.11
0.14
0.06
0.05
0.05
0.04
0.04
0.06
0.26
0.17
0.18
0.17
0.12
0.09
0.12
0.09
0.10
0.07
0.17
0.13
0.15
0.06
0.13
0.11
0.11
0.08
0.09
0.07
0.13
0.12
0.11
0.08
0.08
0.09
0.09
0.08
0.08
0.08
0.19
0.22
0.12
0.15
0.07
0.06
0.06
0.06
0.06
0.07
0.28
0.20
0.21
0.20
0.16
0.13
0.16
0.13
0.14
0.12
Random Uncertainties from Autosynth
M15
M92
M53
N5466
M13
M2
M3
M5
M107
M71
0.11
0.10
0.06
0.06
0.05
0.04
0.04
0.04
0.04
0.02
0.21
0.18
0.16
0.18
0.12
0.11
0.12
0.08
0.04
0.03
0.30
0.30
0.12
0.15
0.06
0.06
0.08
0.05
0.03
0.04
0.06
0.05
0.05
0.03
0.05
0.04
0.05
0.05
0.05
0.04
0.08
0.07
0.06
0.07
0.05
0.07
0.03
0.04
0.04
0.04
Final Combined Uncertainties
M15
M92
M53
N5466
M13
M2
M3
M5
M107
M71
0.12
0.11
0.08
0.08
0.07
0.06
0.06
0.06
0.06
0.05
0.22
0.20
0.18
0.20
0.14
0.14
0.14
0.11
0.09
0.09
0.32
0.32
0.16
0.19
0.13
0.13
0.14
0.12
0.11
0.12
0.13
0.13
0.13
0.12
0.13
0.13
0.13
0.13
0.13
0.13
0.09
0.09
0.07
0.08
0.06
0.08
0.04
0.05
0.05
0.05
Note. Top panel: systematic uncertainty estimates from changes in T eff , log g , and vmicro. Middle panel: average random uncertainties reported by autosynth.
Bottom panel: the final uncertainties are the sum of uncertainties in quadrature from the middle panel and uncertainties for ΔT eff = 100 K, D log g = 0.3, and
Dvmicro = 0.1 km s−1.
two spectra slightly different from the original spectrum. Then,
the fit of each line is repeated while keeping all other parameters
unchanged. The average difference between these two new fits
and the original best-fit spectrum is the defined uncertainty
associated with the fit itself. This uncertainty estimate is mainly
sensitive to variations of noise in the spectrum in the defined
windows for the c 2 fit. If its value is close to 0, then the
(1 + s ) ´ spectrum and (1 - s ) ´ spectrum give very similar
or the same abundances. This is expected when working at high
S/N and with a well defined continuum.
While this uncertainty estimation method is reliable in most
cases, it has its limitations for very noisy spectra, very weak lines,
or when abundances are near upper limits. Uncertainties are
usually overestimated for noisy spectra, while they are underestimated for weak lines and upper limits. Thus, we decided to
use one uncertainty from autosynth per element per cluster by
simply averaging together all uncertainties reported by the
program. This resulted in one uncertainty estimate for every
element in each cluster. Autosynth uncertainties are listed in the
middle section of Table 4. The final estimated uncertainties were
calculated by adding together in quadrature the uncertainties for
100 K difference in temperature (also 0.3 dex systematic
difference in log g, and 0.1 km s−1 in microturbulent velocity),
and the average autosynth values that estimate random
uncertainties. These final estimations for each element per cluster
are given in the bottom panel of Table 4.
4. LITERATURE COMPARISONS
We take [X/Fe] abundance values for C, N, O, Mg, Al, Si,
Ca, and Ti from high-resolution spectroscopic studies in the
literature as a point of comparison for our abundance
determinations. We use the same literature sources as Mészáros
et al. (2013), and added more recently published papers listed
in Table 5. The derivation of stellar parameters Teff and log g
are described in detail in Section 2 and they were also
compared to the literature. Our goal for these comparisons is
not to cross-calibrate our new abundance determinations with
the literature; rather, we are looking for cases where our
abundances are systematically different from the literature or
particular clusters or elements for which our homogeneously
observed and analyzed data set can clarify conflicts in the
literature.
Cross-identification between the GC stars in the DR10
APOGEE data set and the literature was performed using the
Simbad online service22, based on 2MASS identifiers and (α,
δ) coordinates. Because there is a large and heterogeneous
literature on chemical abundances in GC stars, we are
providing our cross-identifications as a resource for the
community. Table 5 lists 2MASS ID and position and alternate
22
8
http://simbad.u-strasbg.fr/simbad/
2MASS ID
2M21290843+1209118
2M21293353+1204552
2M21293871+1211530
2M21294979+1211058
2M21294979+1211058
Cluster
T eff
log g
[Fe/H]
[C/Fe]
[N/Fe]
[O/Fe]
[Mg/Fe]
[Al/Fe]
[Si/Fe]
[Ca/Fe]
[Ti/Fe]
M15
M15
M15
M15
M15
K
K
K
K
K
K
K
K
K
K
−2.37
−2.37
−2.37
−2.40
−2.27
K
K
K
K
K
K
K
K
K
K
K
K
K
K
0.42
K
K
K
K
K
K
K
K
K
K
0.36
K
K
0.53
K
K
0.13
0.14
0.28
0.38
0.31
K
K
K
K
Alt. ID
B5
K22
K47
K144
K144
Source
The Astronomical Journal, 149:153 (24pp), 2015 May
Table 5
Identifiers, Stellar Parameters, and Elemental Abundances from the Literature for Stars in Our Sample
h
h
h
h
i
9
Note. Individual star references: (a) O’Connell et al. (2011); (b) Carretta et al. (2009a); (c) Johnson & Pilachowski (2012); (d) Sneden et al. (2004); (e) Yong et al. (2006); (f) Cohen & Meléndez (2005); (g) Cavallo
& Nagar (2000); (h) Sneden et al. (2000); (i) Minniti et al. (1996); (j) Otsuki et al. (2006); (k) Sneden et al. (1991); (l) Sneden et al. (1997); (m) Sobeck et al. (2011); (n) Kraft & Ivans (2003); (o) Kraft et al. (1992);
(p) Johnson et al. (2005); (q) Lai et al. (2011); (r) Ivans et al. (2001); (s) Koch & McWilliam (2010); (t) Sneden et al. (1992); (u) Ramírez and Cohen (2003); (v) Yong et al. (2008); (w) Meléndez & Cohen (2009);
(x) Briley et al. (1997); (y) Shetrone (1996); (z) Smith et al. (2007); (aa) Lee et al. (2004); (ab) Ramírez & Cohen (2002); (ac) Yong et al. (2006); (ad) Sneden et al. (2000); (ae) Roederer & Sneden (2011).
(This table is available in its entirety in machine-readable and Virtual Observatory (VO) forms.)
Mészáros et al.
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
the more recent solar abundance values from Asplund et al.
(2005) whereas our earlier literature sources do not, we expect
the ∼0.3 dex offset visible in Figure 4. The systematic
difference in [N/Fe] abundance is also likely to be due to the
updated solar abundances.
Carbon is the only element studied in this paper for which
this explanation does not hold: although the solar abundance
was revised down, from A(C)  = 8.56 (Anders &
Grevesse 1989) to 8.39 (Asplund et al. 2005), our [C/Fe]
values are typically lower than the literature values. Unfortunately, literature values are available for only a subset of our
stars, which makes it difficult to verify precisely any systematic
behavior.
In the other elements we consider in the present study, there
is generally good agreement between our results and those from
the literature, which is encouraging. However, we find
significant differences in Al abundances in stars with belowsolar metallicities. This is mainly driven by stars in M3, where
the [Al/Fe] abundances from Johnson et al. (2005) are larger
than in the other literature sources. We also see significant
scatter in [Ca/Fe] and [Ti/Fe] at low abundances; this is likely
caused by our abundance determination method having
difficulty with weak lines.
The [Mg/Fe] versus [Al/Fe] relations in M3, M13, and M5
provide useful examples of how our new data set compares
with the literature. In M3, our abundance determinations show
a clear bimodality in the Mg–Al distribution. The study of
Cavallo & Nagar (2000) found a similar bimodality, but
Johnson et al. (2005) found a smooth distribution. In M13 we
find an extended anticorrelation between [Mg/Fe] and [Al/Fe],
and the literature abundances lie within it. We have seven stars
in common with Sneden et al. (2004), and they span the full
anticorrelation, while the three and two stars, respectively,
which we have in common with Cohen & Meléndez (2005)
and Cavallo & Nagar (2000) are consistent with our abundance
results but happen to inhabit small regions of abundance space.
M5 is a similar case, where our [Al/Fe] values span a range of
−0.3 to +1.1 dex, while the Carretta et al. (2009a) Al
abundances are limited to between −0.2 and +0.7 dex. As in
M13, our sample is larger than the literature sample, and covers
the full RGB, while the literature sample does not fully span the
parameter space.
Figure 3. Differences in abundances produced by two runs adopting different
temperatures: photometric and ASPCAP temperatures; otherwise the same
calculation method was used. The points are color-coded by the differences
between the photometric and ASPCAP temperatures. The ± errors give the
standard deviation around the mean of the differences.
stellar identifiers from literature abundance studies for all of the
stars considered in this study.
Figure 4 shows comparisons of our stellar parameters and
abundances against the literature values. Different GCs are
represented by different colored points. In general, we find a
systematic offset of ∼100 K between our photometric Teff
values and the spectroscopic effective temperatures from the
literature, with a reasonably small scatter. This is similar to the
typical difference between spectroscopic and photometric
temperatures reported by Mészáros et al. (2013). Because of
the degeneracies in deriving stellar parameters, the slightly
higher temperatures in the literature are accompanied by a
systematic offset of ∼+0.2 dex in log g.
There are a few systematic differences between our
abundance results and those in the literature. These can mainly
be traced back to a change in the solar abundance scale as
derived by Asplund et al. (2005). As one example, the [Fe/H]
metallicities we derive are typically 0.1 dex higher than those
from the literature, which is quite similar in magnitude to the
change in A(Fe)  from 7.52 (Anders & Grevesse 1989) to
7.45 (Asplund et al. 2005). Also, the solar abundance of
oxygen was revised significantly, from A(O)  = 8.93 (Anders
& Grevesse 1989) to 8.66 (Asplund et al. 2005). Since we use
5. VARIATIONS IN INDIVIDUAL
ELEMENT ABUNDANCES
Although there are well-known abundance patterns within
GCs, large homogeneous studies that include a wide range of
abundances are rare. Since we can determine abundances of
most of the light elements for the stars in our sample, we
examine the behavior of the known abundance patterns
across a wide range in cluster metallicity and search for
unexpected variations in the α-elements. In this section we
focus on the range of abundance within each cluster, to
separate real abundance variations, bimodalities, and trends
from possible measurement errors. Table 6 lists the average
and standard deviation for each elemental abundance in each
cluster.
5.1. Correlations with T eff
In Figures 5–7 we show all nine derived abundances as a
function of effective temperature. From Figure 5 we conclude
10
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Table 6
Abundance Averages and Scatter
Cluster
[Fe/H]
[C/Fe]
[N/Fe]
[O/Fe]
A(C+N+O)
[Mg/Fe]
[Al/Fe]
[Si/Fe]
[Ca/Fe]
[Ti/Fe]
0.11
0.14
0.11
0.14
0.13
0.26
0.15
0.23
0.24
0.38
0.34
0.42
0.37
−0.24
0.61
0.45
0.21
0.36
0.47
0.51
0.44
0.45
0.41
0.29
0.40
0.35
0.30
0.34
0.48
0.39
0.16
0.10
0.23
0.04
0.26
0.24
0.12
0.20
0.15
0.21
0.15
0.09
0.28
0.29
0.20
0.27
0.11
0.26
0.21
0.42
0.24
0.23
0.08
0.06
0.15
0.07
0.06
0.09
0.10a
0.05a
0.52
0.48
0.51
0.35
0.53
0.51
0.43
0.34
0.15
0.08
0.16
0.12
0.05
0.09
0.09
0.06
0.04
0.07
0.05
0.05
0.25
0.17
0.17
0.25
0.14
0.12
0.08
0.11
0.21
0.10
0.08
0.20
0.13
0.14
0.14
0.08a
0.13
0.12a
0.16
0.09
Averages
M15
M92
M53
N5466
M13
M2
M3
M5
M107
M71
−2.28
−2.23
−1.95
−1.82
−1.50
−1.49
−1.40
−1.24
−1.01
−0.68
−0.41
−0.41
−0.50
−0.56
−0.53
−0.48
−0.46
−0.46
−0.21
−0.10
0.95
0.93
1.06
0.84
0.89
0.90
0.69
0.76
0.69
0.91
0.54
0.58
0.56
0.63
0.28
0.41
0.40
0.27
0.33
0.51
7.09
7.19
7.49
7.60
7.69
7.76
7.84
7.85
8.15
8.65
Scatter
M15
M92
M53
N5466
M13
M2
M3
M5
M107
M71
a
0.10
0.10
0.07
0.08
0.07
0.08
0.08
0.08
0.06
0.07
0.13
0.11
0.16
0.01
0.07a
0.05a
0.08a
0.13a
0.09
0.13
0.35
0.23
0.21
0.10
0.17
0.15
0.13
0.17
0.27
0.32
0.19
0.19
0.06
0.04
0.17
0.21
0.23
0.27
0.15
0.09
0.14
0.13
0.09
0.08
0.15
0.23
0.15
0.19
0.15
0.13
Standard deviation around the linear fit.
Table 7
Averages of Populations
Cluster
M15
M92
M53
N5466
M13
M2
M3
M5
M107d
M71d
Pop.a
Nb
Rc
[Fe/H]
[C/Fe]
[N/Fe]
[O/Fe]
[Mg/Fe]
[Al/Fe]
[Si/Fe]
[Ca/Fe]
[Ti/Fe]
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
10
13
14
33
8
8
6
2
32
49
7
11
39
20
60
62
6
6
3
4
0.43
0.57
0.30
0.70
0.50
0.50
K
K
0.40
0.60
0.39
0.61
0.66
0.34
0.49
0.51
K
K
K
K
−2.31
−2.26
−2.23
−2.23
−1.96
−1.95
−1.83
−1.79
−1.50
−1.49
−1.46
−1.50
−1.40
−1.38
−1.24
−1.24
−1.04
−1.00
−0.72
−0.69
−0.49
−0.33
−0.42
−0.38
−0.43
−0.54
−0.56
−0.56
−0.44
−0.58
−0.45
−0.49
−0.42
−0.54
−0.41
−0.50
−0.16
−0.25
0.00
−0.17
0.78
1.11
0.89
1.03
0.91
1.13
0.84
0.86
0.83
0.92
0.75
1.01
0.59
0.86
0.62
0.87
0.47
0.92
0.57
1.16
0.64
0.43
0.67
0.40
0.59
0.55
0.64
0.62
0.56
0.12
0.56
0.30
0.53
0.16
0.41
0.16
0.38
0.29
0.53
0.49
0.20
0.04
0.33
0.05
0.38
0.25
0.15
0.11
0.21
0.09
0.30
0.24
0.18
0.10
0.24
0.21
0.36
0.37
0.46
0.44
−0.19
0.75
−0.23
0.70
−0.11
0.84
−0.42
0.31
0.04
0.99
−0.10
0.79
−0.08
0.79
0.06
0.64
0.51
0.53
0.44
0.53
0.31
0.53
0.41
0.47
0.42
0.41
0.28
0.30
0.38
0.40
0.35
0.35
0.30
0.29
0.34
0.35
0.50
0.45
0.41
0.38
0.08
0.24
0.08
0.11
0.23
0.23
0.14
−0.16
0.24
0.28
0.19
0.27
0.13
0.10
0.20
0.20
0.21
0.22
0.24
0.21
0.11
0.17
0.09
0.09
0.24
0.31
0.29
K
0.18
0.21
0.26
0.27
0.11
0.12
0.24
0.28
0.12
0.16
0.38
0.45
a
Population 1 is denoted by red, while population 2 is denoted by blue in Figures 8–12.
N: the number of stars in each population.
c
R: the ratio of number of stars in a population and the overall number of stars analyzed.
d
Populations are found using N instead of Al.
b
that we measure constant Fe abundances in all of the clusters.
Mg and Al abundances show a large range of values in some
clusters (as discussed in Section 2, and further in Section 6),
but no significant trends with T eff except in M107 and M71.
This trend is very weak in M71, and data are consistent with
showing no trend within the uncertainties. However, in M107
the trend is stronger. We currently do not fully understand
where these small correlations come from. We suspect that this
11
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Figure 4. T eff , log g , [Fe/H], [C/Fe], [N/Fe], [Mg/Fe], [Al/Fe], [Si/Fe], [Ca/Fe], [Ti/Fe] compared to literature sources. Different colors denote different clusters; for an
explanation of the colors see upper right panel. Solid lines show the 1:1 relation, while dashed lines denote ±100 K for T eff , and ±0.2 dex for log g and individual
abundances. A detailed discussion can be found in Section 4.
12
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Table 8
Statistics of Correlations
Cluster
M15
M92
M53
M13
M2
M3
M5
a
b
c
Δa
[Mg/Fe]
eΔb
[Mg/Fe]
σc
Δ
[Si/Fe]
eΔ
[Si/Fe]
σ
0.16
0.28
0.13
0.12
0.06
0.08
0.03
0.037
0.029
0.035
0.014
0.038
0.011
0.008
4.3
9.7
3.7
8.6
1.6
7.3
3.8
0.22
0.06
0.01
0.02
0.00
0.01
0.01
0.055
0.038
0.055
0.018
0.043
0.024
0.014
4.0
1.6
<1
1.1
<1
<1
<1
large errors of N, one also has to be careful with possible
correlations of C and N with temperature because they can be
generated on the RGB as part of deep mixing and also in a
previous stellar evolution event that are responsible for the
pollution.
In Figure 7 there is a weak trend, on par with the error,
visible in the Si abundance in M13, but not in the other
clusters. While Ca does not show any correlation with
temperature, it does show temperature-dependent scatter in
most of the GCs. The three Ca lines used in our analysis are
generally weak and get significantly weaker above 4700 K,
which leads to higher errors related to increased sensitivity to
the uncertainty in the continuum placement. Titanium,
unexpectedly, shows a decline with decreasing temperature in
M2, M5 and (marginally) in M107 in Figure 7. We suspect that
inaccuracies in our analysis of the hotter star spectra are driving
this apparent trend since the S/N is lower for those stars than
for the cooler, brighter giants. Because these trends in Mg
(Figure 5), Si, and Ti (Figure 7) only show up in a handful of
clusters, we believe that they are a result of difficulties in data
analysis for certain lines in certain stars and not any systematic
mishandling in, e.g., vmicro estimation.
We choose to fit these various observed trends with a linear
equation, regardless of their origin, in order to explore the
scatter around the trend. We fit lines to [Mg/Fe] in M107 and
M71; [C/Fe] in M13, M2, M3, and M5 and [Ti/Fe] in M2 and
M5. We are only using these fits to determine the internal
scatter, which we define as the standard deviation about the
fitted line. Throughout the rest of this paper we use the
abundance values directly, except when discussing the CN
anticorrelations in Section 6.4.
Difference of the average Mg and Si abundance of populations.
The estimated error of the difference.
Detection significance in σ.
is a result of a combination of effects. One such effect may be
the use of model atmospheres that assume LTE, but that nonLTE effects (Bergemann & Nordlander 2014) act to make the
strong Mg lines in these metal-rich stars give the appearance of
higher abundance in the cooler stars. Other possible effects are
small systematic errors from estimating Teff and log g
culminating during the synthesis. Abundances are also
sensitive to the microturbulent velosity, so if the log g −vmicro
expression used is not as accurate in this metallicity range, or
the surface gravity is badly estimated, this propagates into
abundances that are systematically off through vmicro.
The minimum [C/Fe] (Figure 6) that can be measured from
the CO lines strongly depends on temperature. Because RGB
stars in GCs generally have low carbon abundances, we can
only set upper limits for a number of our stars. Our [C/Fe]
values are an average of the derived abundances from five CO
windows. As a result determining the upper limit is more
challenging because CO lines in certain windows disappear
faster with rising temperature than in others. We carefully
checked every CO window fit and selected upper limits if the
CO band head was not visible in more than three windows by
looking at the flatness of the c 2 fit around the minimum value.
Because the derived abundance of N from the CN lines is
anticorrelated with the value of [C/Fe] used, all stars with upper
limits in [C/Fe] have also lower limits in [N/Fe]. These upper
limits for [C/Fe] and lower limits for [N/Fe] are identified with
open triangles in all figures.
We see clear correlations between [C/Fe] and T eff in M13,
M3, M2, and M5 when omitting upper limits. We interpret
these as a sign of “deep mixing,” a nonconvective mixing
process that causes steady depletion of surface carbon
abundance and an enhancement in nitrogen abundance in all
low-mass RGB stars (Gratton et al. 2004). We derive only
upper limits on [C/Fe] for M15, M92, M53, and NGC 5466
because the CO bands are quite weak at such low metallicity.
With only upper limits, it is not possible to identify abundance
trends in these clusters. However, previous studies (Martell
et al. 2008; Shetrone et al. 2010; Angelou et al. 2012) have
found signs of the same deep mixing process in action in those
clusters.
The only cluster in which we see a sign of nitrogen
enrichment with declining temperature is M13. Possible trends
in nitrogen are not necessarily a result of astrophysics, because
in our methodology small systematic errors in temperature can
result in systematic errors in N abundance because deriving [N/
Fe] is challenging from the CN lines. Other than the relatively
5.2. Scatter and Errors
An internal abundance scatter significantly larger than the
estimated errors suggests an astrophysical origin. In fact,
significant scatter in N, O, Mg, and Al is well documented in
the literature (Gratton et al. 2012). The large star-to-star
variations in these elements are the result of the CNO, Ne–Na,
and Mg–Al cycles, which play an important role in the nuclear
fusion processes at the early stages of RGB (CNO) and AGB
(Ne–Na, Mg–Al) evolution. Figure 8 shows the internal scatter
calculated as the standard deviation around the mean values (in
some cases around a fitted linear equation, as explained
earlier), and the final combined estimated uncertainties from
Table 4, as a function of cluster average metallicity. We note
that the small number of stars analyzed in NGC 5466 limits our
ability to make a detailed analysis of C, N, and O in this cluster:
because we were only able to measure the abundance of these
elements in the three stars that have temperatures below
4500 K, the scatter is unrealistically higher than the calculated
uncertainty.
The star-to-star scatter in [Fe/H] is quite similar to our
measurement uncertainties, indicating that no significant
metallicity variations in the clusters are detected in this study.
Recently, Yong et al. (2014) have discovered three distinctive
groups with different iron abundances in M2. The first group
has a dominant peak at [Fe/H] = −1.7, while the second and
third have smaller peaks at 1.5 and 1.0, but the membership for
the latter group is not conclusive. We see no such behavior in
M2. The discrepancy between these two studies is most likely
the result of a selection effect. Yong et al. (2014) selected stars
that belong to a second RGB (Lardo et al. 2012) within the
13
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Figure 5. [Fe/H], [Mg/Fe], and [Al/Fe] as a function of photometric T eff for all 10 clusters. The error bars represent our final combined uncertainties from Table 4. The
linear fit is plotted over [Mg/Fe] for M107 and M71 to remove the visible linear trend (see Section 5.1 for discussion).
Martell et al. (2008), the rate of carbon depletion due to deep
mixing is lower in higher-metallicity stars.
The range in [N/Fe] abundance (Figure 8, second panel from
the top, in the left column) in M71, M107, M5, M3, M13, and
M2 is clearly larger than the estimated uncertainties. However,
the difference between the scatter and the uncertainty decreases
with decreasing cluster metallicity, and below [Fe/H] = −1.7
(NGC 5466, M53, M92, and M15) our derived [N/Fe] values are
all lower limits. The decreasing discrepancy is a result of larger
uncertainties in more metal-poor clusters and the fact that the
minimum N abundance of a cluster is slightly increasing as the
average metallicity decreases (Figure 6), while the maximum N
is constant above [M/H] = −1.7. The uncertainties in M15 and
M92 are significantly larger than in other clusters, and our
measurements are not precise enough to judge the amount of N
enrichment in these clusters. Fitting the CN lines in our defined
CN window is also challenging. This is because they are usually
blended with other molecular lines and because N abundance
strongly depends on the derived C abundance, which is an upper
limit in many cases. A more detailed analysis of the C–N
anticorrelation and N bimodality can be found in Section 6.4.
Our uncertainties of the [O/Fe] abundance (Figure 8) are
fairly constant as a function of cluster metallicity. This is
because the error is dominated by a strong sensitivity to the
temperature used in the model atmosphere, and the internal
scatter varies between clusters. Unfortunately, the Na lines
cluster, while our sample selection was based upon previous
observations (Zasowski et al. 2013) and all of our stars belong
to the main RGB of this cluster.
The α-elements Si and Ti also show constant distribution,
but the scatter in [Ca/Fe] is larger than our measurement
uncertainties for most clusters in our sample. We attribute this
to an increasing inaccuracy in [Ca/Fe] with rising temperature,
as discussed in Section 5.1, and not a real range in calcium
abundance. Little to no range in iron and α-elements is what we
expect for normal GCs, although the clusters in which these
abundances do vary are an intriguing puzzle (see, e.g., Marino
et al. 2013). M107 stands out as having the largest discrepancy
between the scatter of [Ca/Fe] and its estimated uncertainty.
However, the large internal scatter is not related to astrophysics, but to spectral contamination from telluric OH lines.
Due to the cluster’s radial velocity, two of the three Ca lines lie
on top of atmospheric OH lines. Because the telluric correction
is not perfect, the Ca lines are compromised in a way that is not
included in our uncertainty estimation.
We do see scatter above the level of the uncertainties in
some of the light-element abundances. In the case of carbon,
we find that the scatter around the overall carbon depletion
trend for M13, M2, M3, and M5 (as shown by dashed lines in
Figure 6) is fairly small, although our measurement uncertainties get quite large at low metallicity. We do not see clear
carbon depletion in M107 or M71; however, as discussed in
14
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Figure 6. [C/Fe], [N/Fe], and [O/Fe] as a function of photometric T eff for all 10 clusters. Open triangles mark upper limits for [C/Fe] and lower limits for [N/Fe], while
the real detections are plotted using filled red dots. The error bars represent our final combined uncertainties from Table 4. The linear correlation in [C/Fe] as a function
of T eff in M13, M2, M3, and M5 is the effect of CNO burning on the RGB. The fitted lines are used to remove the trend in order to estimate the scatter in these clusters
(see Section 5.1 for discussion).
populations in more detail in the next section. The derived αelement abundances are also fairly constant in all GCs presented
here, similar to what is reported in the literature (Gratton
et al. 2004).
available in the APOGEE spectra are too weak to investigate
the O–Na anticorrelation in our sample, so we can only limit
our discussion to the scatter of [O/Fe]. We use the latter as an
indication of the strength of O–Na anticorrelation in a cluster.
In M71 and M107 the O scatter is comparable to the
uncertainties, suggesting that the O–Na anticorrelation is not
as extended as in the rest of the clusters. A clear spread is
visible in M5, M3, M2, M13, M15, and M92, while in M53 the
O abundances are constant and the scatter is significantly
smaller than what is expected from the uncertainties.
The effects of the Mg–Al cycle are very noticeable for all
clusters except M71 and M107 (Figure 8, second panels from
the bottom). The scatter in Mg abundance in those clusters is
close to the estimated uncertainties after taking the linear
correlations with T eff into account (Figure 5). In other clusters
the scatter of [Mg/Fe] also closely follows the uncertainties,
except for M15 and M92 where the most Mg poor stars can be
found. We see no Al enrichment in M107 and M71; however,
as the average metallicity decreases, the amount of internal
scatter rapidly starts to deviate from the error, and we see a
large spread in Al in all other GCs. A more detailed analysis of
the Mg–Al anticorrelations and Al populations can be found in
Sections 6.2 and 6.3.
Based on the examination of the spread of N and Al
abundances, we conclude that all clusters possess multiple stellar
populations either based on N or Al, or both. We examine these
6. MULTIPLE POPULATIONS
Most light elements show star-to-star variations in all GCs.
These large variations are generally interpreted as the result of
chemical feedback from an earlier generation of stars (Gratton
et al. 2001; Cohen et al. 2002), rather than inhomogeneities in
the original stellar cloud from which these stars formed. Thus,
the current scenario of GC evolution generally assumes that
more than one population of stars was formed in each cluster.
The first generation of GC stars formed from gas that had been
enriched by supernovae in the very early universe, while the
second formed by combining gas from the original star-forming
cloud with ejecta from the first generation stars. Only a fraction
of first generation stars contribute to the pollution, and the
timescale of the formation of these second generation stars
depend on the nature of polluters; it is a couple of hundred Myr
in the case of intermediate-mass AGB stars, but it is only a few
Myr for fast rotating massive stars and massive binaries.
These first generation pollutors are thought to have been fast
rotating massive stars (Decressin et al. 2007, intermediate mass
(Mstar > 3 M) AGB stars (D’Ercole et al. 2008), or massive
15
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Figure 7. [Si/Fe], [Ca/Fe], and [Ti/Fe] as a function of photometric T eff for all 10 clusters. The error bars represent our final combined uncertainties from Table 4.
Trends in Ti for M2 and M5 are removed using the plotted lines when calculating the internal scatter (see Section 5.1 for discussion).
binaries (de Mink et al. 2009). Our data reveal the expected
signatures of pollution from material enriched from the hot
hydrogen burning cycles such as the CNO, NeNa, or MgAl
cycles in all GCs in our sample. In this section we explore the
various correlations between these elements, and we discuss
individual star formation, and/or pollution events suggested by
separate groups found in the C–N and Mg–Al anticorrelations.
algorithm; XD itself does not determine the optimal number
of components to fit the distribution.
Briefly, the XD algorithm works by optimizing the likelihood of the Gaussian mixture model of the data. The
optimization proceeds by an iterative procedure consisting of
repeated expectation (E) and maximization (M) steps. In the E
step, each datum is probabilistically assigned (a) membership
in each population and (b) error-free values of each noisy or
missing elemental abundance; in the M step, each Gaussian’s
parameters are updated using its members’ mean abundances
and covariance calculated from the error-free values obtained in
the E step. These steps are repeated until the likelihood stops
increasing to within a small tolerance. The algorithm is proven
to increase the likelihood in each EM-step.
We analyze each GC in our sample using this algorithm,
with the [Mg/Fe], [Al/Fe], [Si/Fe], [Ca/Fe], [Ti/Fe] abundances
as well as only [Mg/Fe] and [Al/Fe]. We find that the two
separate analyses provide nearly identical results, indicating
that Si, Ca, and Ti do not drive the populations. For M71 and
M107 we used [N/Fe] instead of [Al/Fe], because Al does not
show any spread in these two clusters. We run XD using
uncertainties given by the “final combined errors” section in
Table 4, and also without any uncertainties, and found that the
two different runs produce identical results. We include
missing abundance measurements by using a large uncertainty
for these measurements. Group membership for each star is
6.1. Identifying Multiple Populations
To separate the various populations under study, we follow
an approach similar to that of Gratton et al. (2011), who used
K-means clustering (Steinhaus 1956) to identify multiple
populations in ωCen. Here, we use the extreme-deconvolution
(XD) method of Bovy et al. (2011)23 to identify population
groups and assign membership. This method fits the
distribution of a vector quantity— here the elemental
abundances— as a sum of K Gaussian populations, whose
amplitudes, centers, and covariance matrices are left entirely
free. The algorithm can be applied to noisy or incomplete
data, thus making the best possible use of all available data.
In the current application, XD’s main advantage is the latter
since we are not able to measure abundances for all nine
elements in all stars in our sample. Similar to the K-means,
the number K of populations to fit is an input to the
23
Code available at http://github.com/jobovy/extreme-deconvolution.
16
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Figure 8. Internal scatter in each cluster from Table 6 (red dots) compared to our estimated final combined errors from Table 4 (blue dots).
determined using the best-fit Gaussian mixture by calculating
the posterior probability for each star to be a member of each
population based on its elemental abundances and their
uncertainties; stars are then assigned to the population for
which this probability is the largest. These membership
assignments are typically unambiguous (probabilities 99%
in most cases), with only a few cases for which the maximum
probability is below 90%. The abundance averages of
populations are listed in Table 7. In the next few subsections
we discuss the results from the population fitting in more detail
for each individual cluster.
blue points in the panels for M107 and M71 are stars that are
warmer than 4500 K, and thus have no CNO measurements.
The extended distribution of Mg and Al abundances
apparent in Figure 9 is typical for GCs, and it shows the
influence of the Mg–Al fusion cycle, which converts Mg into
Al. However, there is a variety in the structure of this
relationship that has not been thoroughly explored before: in
some clusters (M92, M53, NGC 5466, M2, and M3), there are
two distinct abundance groups with a gap, while in other
clusters (M15, M13, and M5) there is no gap. This result
strongly suggests that there is diversity in the process of stellar
chemical feedback and star formation in GCs, which may relate
to the larger environment in which they formed.
To date, the largest homogeneous study of Mg and Al
abundances was carried out by Carretta et al. (2009a). We have
several clusters in common, M15, M5, M107, and M71 which
enables a direct comparison. M107 and M71 do not show any
anticorrelation in either of the studies. Carretta et al. (2009a)
only have upper limits for Al and Mg in M15, while we were
able to make direct measurements, but even with upper limits
the two studies show similar anticorrelations with largest
spreads in both Al and Mg from the whole sample of clusters.
The largest difference between the two studies can be seen in
case of M5, where our values span a range of −0.3 to +1.1 dex,
while the Al abundances of Carretta et al. (2009a) are limited
to between −0.2 and 0.7 dex. The difference may be explained
by a simple selection effect, as Carretta et al. (2009a) observed
6.2. The Mg–Al Anticorrelation
In order for the Mg–Al cycle to operate, high temperatures
above 70 million K are required (Charbonnel & Prantzos 2006).
Because current cluster main-sequence stars are unable to reach
these temperatures, the high [Al/Fe] abundances we see in some
GC stars imply that a previous generation of higher-mass or
evolved stars must have contributed to their chemical
composition.
Figure 9 shows [Mg/Fe] versus [Al/Fe] for all 10 clusters in
our study. The two XD populations are plotted in red and blue,
and a representative error bar is included in the top right corner
of each panel. As noted previously, stars in M107 and M71 do
not show a strong Mg–Al anticorrelation, and their XD
population assignments are based on N abundances. The light
17
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Figure 9. Mg–Al anticorrelations. Colors mark the populations found with the XD code; the first population is denoted by red, and the second-generation denoted by
blue points. Because [N/Fe] was used to identify populations in M107 and M71, not all stars could be cataloged; these are denoted by light blue. The same colors
denote the same stellar groupings in Figures 9–13.
13 stars in M5, while our sample size is 122 and covers almost
the full extent of the RGB.
We discuss the extent of anticorrelation through differences
in the average of the Mg abundance between the populations,
divided by the estimated uncertainty in this average Mg
difference. We use the final combined errors from Table 4.
Table 8 lists the errors of the average differences with values of
how much σ detection is each difference. We find that in M15
(4.3σ), M92 (9.7σ), M13 (8.6σ), and M3 (7.3σ) the
differences are statistically significant and the Mg–Al anticorrelation exists. Two clusters, M53 (3.7σ) and M5 (3.8σ),
also show large σ detections, however, we would like to use a
more conservative approach to the detection of anticorrelations
out of a concern that our errors may be underestimated, given
the degeneracies between stellar parameters and abundances in
our analysis. M2 stands out as having no statistically significant
Mg–Al anticorrelation.
The summed abundance A(Mg+Al) is expected to be
constant as a function of T eff when material is completely
processed through the Mg–Al cycle, and that is what our results
show in Figure 10. The only exception is M107, but the slight
correlation is due to Mg correlating with temperature (see
Figure 5), and this trend was not removed here.
Al is expected to correlate with elements enhanced by
proton-capture reactions (N, Na; Figure 11) and anti-
correlate with those depleted in H-burning at high temperature (O, Mg; Figure 11). The Al–O anticorrelations and Mg–
O correlations can be clearly seen in our data in Figure 11 for
all clusters except M53, M107, and M71. The small number
of stars observed in NGC 5466 does not allow us to
investigate the correlations in that cluster in detail. However,
the slight correlation of Al with Si that can be seen in M15 in
Figure 11 is the evidence of 28Si leaking from the Mg–Al
cycle, which is an intriguing result. Si participating in the
light-element abundance pattern was first reported in
NGC 6752 by Yong et al. (2005). Since then, Carretta
et al. (2009a) have also found Si enhancement correlated
with Al in NGC 2808. Those authors’ interpretation was that
it is only in low-metallilcity clusters, where the AGB stars
burn slightly hotter, or in high-mass clusters, where the
chemical enrichment is more efficient, that a Si enhancement
will be observed in second-generation GC stars. The
difference in Si abundance between the two XD-identified
populations, listed in Table 8, is significant relative to the
error on that measurement. The Si–Al correlation in M15 is
also accompanied by a Si–Mg anticorrelation (Figure 11),
which is further evidence of 28Si being produced by hot
bottom burning (HBB) in AGB stars (Karakas & Lattanzio 2003).
18
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Figure 10. Combined abundance of A(Mg+Al) as a function of effective temperature.
lower than their solar metallicity counterparts (M > 4 M)
(García-Hernández et al. 2006, 2009). Thus, at the lowest
metallicities in M15 and M92 we also would expect more HBB
AGB stars (i.e., with several degrees of HBB and Mg–Al
nucleosynthesis), because the minimum stellar mass to activate
the HBB process (and advanced Mg–Al nucleosynthesis)
decreases with decreasing metallicity. Thus, we conclude that
more polluted material would be present at the lowest
metallicities.
Because of these reasons, we believe that our results support
and add some evidence to the high-mass AGBs as GCs
polluters.
6.3. The Spread of Al Abundances
The spread of Al abundances (Figure 8) also increases
significantly below [Fe/H] = −1.1, thus we can conclude that
high mass, low metallicity AGB polluters that are able to
process material at high, 60–70 million K, play an increasingly
larger role in more metal-poor cluster. This behavior of Al
abundances was previously observed by Carretta et al. (2009a),
but the larger range of Al values presented in this paper makes
this correlation clearer.
Theoretical AGB nucleosynthesis modeling indeed predicts
this behavior (Ventura et al. 2001, 2013). The high-mass AGB
stars reach higher temperatures at the bottom of the convective
envelope; i.e., stronger HBB and advanced (Mg–Al) nucleosynthesis occurs with decreasing metallicity. We are seeing
very advanced Mg–Al nucleosynthesis in the most metal-poor
clusters such as M15 and M92, while the most metal-rich
clusters like M107 and M71 do not show this, as theoretically
expected, if the high-mass AGB stars are the polluters. Also,
this is corroborated by the Al–O anticorrelation; the most Alrich stars are O-poor showing the effects of very strong HBB,
because HBB proceeds completely and destroys O.
An other related issue is that HBB is activated for lower
masses with decreasing metallicity. High-mass and very lowmetallicity AGB stars do not exist in GCs today due to their
extremely short lifetimes, but this trend seems to be confirmed
both theoretically (Ventura et al. 2001, 2013) and observationally
at least for solar metallicities down to those of the Magellanic
Clouds. From the observational point of view, we know that
high-mass AGB stars in the Small Magellanic Cloud ([Fe/
H] = −0.7) activate HBB for progenitor masses (M > 3 M)
6.4. C-N
Carbon, nitrogen, and sometimes oxygen are influenced by
two independent processes in GC giants: a primordial anticorrelation, with the same first generation sources as the O–Na
and Mg–Al patterns and also stellar evolution, driven by
circulation between the hydrogen-burning shell and the surface
(Sweigart & Mengel 1979; Angelou et al. 2012). As discussed
in the previous section, we see large star-to-star variations in N
abundance in all clusters; however, our inability to measure C
abundances below [Fe/H] = −1.7 dex leads us to only having
upper limits, and therefore we restrict the discussion of C–N
anticorrelations to the more metal-rich clusters.
Figure 12 shows [N/Fe] versus [C/Fe] for all clusters. Deep
mixing has to be taken into account when discussing the CN
anticorrelations, so in M13, M2, M3, and M5, where we see
deep mixing clearly, the correlation of [C/Fe] with T eff was
removed. Because deep mixing is not visible in M107 and
19
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Figure 11. Si and O (anti)correlations with Al, N, and Mg. For explanation of color coding, please see description of Figure 9.
20
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Figure 12. CN anticorrelations. Correlations of [C/Fe] with temperature associated with deep mixing were removed in clusters marked by three stars. Upper limits are
denoted by open triangles.
M71, we used the original values. A clear anticorrelation
cannot be seen in any of the clusters. While M107 and M71
show weak anticorrelations, neither of these is statistically
significant. The anticorrelation itself may be obscured by the
relatively large errors for [C/Fe] and [N/Fe]. As mentioned in
Section 5.1, we see correlations of [C/Fe] with temperature in
M13, M3, M2, and M5, but these are not accompanied by
increasing [N/Fe] at the same time. We believe that this is due
to our inability to measure C and N accurately: most of our
abundance determinations are upper limits or close to them. If
the anticorrelations exist, they must span ranges smaller than
our uncertainties, which are always higher than 0.14 dex in [C/
Fe] and 0.12 dex in [N/Fe]. It is interesting to note that in M3,
M107, and M71 there is a clear CN weak and CN strong group
of stars, while in all more metal-poor clusters, N abundances
fill out a broad distribution.
attributed to larger than usual pollution from lower mass AGB
stars than in other clusters, but these results were questioned by
Villanova et al. (2010) as they did not see spread of C+N+O
larger than 0.3 dex.
We also see near constant C+N+O in our sample, which is
consistent with the material in these stars having undergone
CNO cycling in the first-generation of stars in the RGB phase.
In Figure 13 we show A(C+N+O) as a function of T eff . The
spread is significantly larger than what is reported in literature,
between 0.4 and 0.6 dex for M13, M2, M3, and M5. The spread
in these clusters is at the level of what has been found in
N1851, but we think this is mostly associated with large
uncertainties of [N/Fe] measurements. We do not find clear
separation in the amount of A(C+N+O) between the two
populations found in any of the clusters.
As previously mentioned, there are currently two leading
models to explain the nature of the polluters: the first assumes
that intermediate-mass AGB stars that are thermally pulsing
and undergoing HBB expel material to the intra-cluster
medium by strong mass loss (Ventura et al. 2001), while the
second assumes that the pollution comes from hot, fast rotating
stars (Decressin et al. 2007). According to the first theory,
intermediate-mass HBB AGB stars could produce or not
produce large C+N+O variations (see, e.g., Yong et al. 2009
and references therein). This is because the C+N+O predictions
(which basically depend on the number of third dredge-up
6.5. C+N+O
We have C, N, and O abundances available for a large
number of stars, thus we are able to investigate the C+N+O
content in each cluster in detail. Studies from the literature
showed that the C+N+O content in globular clusters are fairly
constant to within 0.3 dex (Smith et al. 1996; Ivans et al. 1999;
Carretta et al. 2005), except for N1851 where Yong et al.
(2009) find a spread of 0.57 dex. According to Yong et al.
(2009) the large spread in C+N+O in N1851 is probably
21
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Figure 13. Upper panels: the sum of C, N, and O as a function of effective temperature. Upper limits are denoted by open triangles. For explanation of color coding,
please see description of Figure 8. Lower panels: the sum of C, N, and O as a function of [Al/H]. A clear correlation is visible in M13, M2, M3, and M5, for more
discussion see Section 6.4.
22
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
episodes) for these stars are extremely dependent on the
theoretical modeling, the convective model, and the mass loss
prescription used (see, e.g., DAntona & Ventura 2008; Karakas
et al. 2012). A (C+N+O)–Al correlation is not clearly visible in
the lower panel of Figure 10.
According to the first theory, AGB stars may explain a large
spread in A(C+N+O) because they are expected to produce a
substantial increase in the C+N+O abundance as they enhance
Na and Al and deplete O and Mg (Yong et al. 2009). This
should result in a (C+N+O)–Al correlation, which is not clearly
visible in the lower panel of Figure 13. Weak correlations in
M107 and M71, and weak anticorrelations in M13 and M3 are
visible, but they are probably the result of large uncertainties in
C and N. More accurate measurements of [N/Fe] will help to
clarify the possible (C+N+O)–Al correlation.
of measuring abundances of CNO. The total abundance
of C+N+O is found to be constant in each cluster, within
our errors. Our data set does not show an unambiguous
correlation between A(C+N+O) and [Al/H]. In principle,
this finding is not against the idea of intermediate-mass
HBB AGB stars as the primary source of chemical selfenrichment in globular clusters. However, more precise
measurements of C and N abundances will clarify this
issue, and we expect that ASPCAP will be able to provide
these in the future.
Szabolcs Mészáros is especially grateful for the technical
expertise and assistance provided by the Department of
Astronomy at Indiana University. We thank Eileen D. Friel,
Catherine A. Pilachowski, and Enrico Vesperini for their
detailed comments during the development of autosynth. Sarah
Martell acknowledges the support of the Australian Research
Council through DECRA Fellowship DE140100598. Sara
Lucatello acknowledges partial support from grant PRIN
MIUR 2010-2011 “Chemical and dynamical evolution of the
Milky Way and Local Group Galaxies.” Jo Bovy was
supported by NASA through Hubble Fellowship grant HSTHF-51285.01 from the Space Telescope Science Institute,
which is operated by the Association of Universities for
Research in Astronomy, Incorporated, under NASA contract
NAS5-26555. Timothy C. Beers acknowledges partial support
for this work from grants PHY 08-22648; Physics Frontier
Center/Joint Institute or Nuclear Astrophysics (JINA), and
PHY 14-30152; Physics Frontier Center/JINA Center for the
Evolution of the Elements (JINA-CEE), awarded by the US
National Science Foundation. Katia Cunha acknowledges
support for this research from the National Science Foundation
(AST-0907873). Verne V. Smith acknowledges partial support
for this research from the National Science Foundation (AST1109888). D. A. García-Hernández and Olga Zamora acknowledge support provided by the Spanish Ministry of Economy
and Competitiveness under grant AYA-2011-27754. Funding
for SDSS-III has been provided by the Alfred P. Sloan
Foundation, the Participating Institutions, the National Science
Foundation, and the U.S. Department of Energy Office of
Science. The SDSS-III website is http://sdss3.org/. SDSS-III is
managed by the Astrophysical Research Consortium for the
Participating Institutions of the SDSS-III Collaboration including the University of Arizona, the Brazilian Participation
Group, Brookhaven National Laboratory, University of Cambridge, Carnegie Mellon University, University of Florida, the
French Participation Group, the German Participation Group,
Harvard University, the Instituto de Astrofisica de Canarias, the
Michigan State/Notre Dame/JINA Participation Group, Johns
Hopkins University, Lawrence Berkeley National Laboratory,
Max Planck Institute for Astrophysics, New Mexico State
University, New York University, Ohio State University,
Pennsylvania State University, University of Portsmouth,
Princeton University, the Spanish Participation Group, University of Tokyo, University of Utah, Vanderbilt University,
University of Virginia, University of Washington, and Yale
University.
7. SUMMARY
We investigated the abundances of nine elements for 428
stars in 10 GCs using APOGEE DR10 spectra. A homogeneous
analysis of these GCs has not been accomplished previously,
something that APOGEE is uniquely able to do because it can
observe all the bright GCs in the northern hemisphere. A
semi-automated code called autosynth was developed to
provide abundances independent of those derived by ASPCAP.
Our main goal was to examine the stellar populations in each
cluster using a homogeneous data set. Based on our
abundances, we find the following:
1. From the examination of the star-to-star scatter in αelement abundance, only O and Mg show an abundance
scatter larger than our measurement errors. However, this
is expected, because O and Mg are part of the wellknown GC light-element abundance anticorrelations.
2. Population analysis using [N/Fe] for M107 and M71 and
[Al/Fe] in the rest of the clusters confirms that each
cluster can be divided into two discrete populations,
though in M107 and M71 that division is clearest in the
N–Al plane, whereas in the other clusters it is best made
in the Mg–Al plane.
3. The anticorrelation of Mg and Al is clearly shown in
M15, M92, M13, and M3, and to a lesser extent in M53
and M5. Interestingly, M2, which is also a metal-poor
cluster, does not have a Mg–Al anticorrelation. The
increased number of stars in our sample compared to the
literature enables us to discover more Al-rich stars,
making the anticorrelation clearer in this data set than in
previous studies.
4. The spread of Al abundances increases with decreasing
cluster metallicity. This is in agreement with theoretical
AGB nucleosynthesis predictions by Ventura et al. (2001,
2013). This suggests that high-mass HBB AGB stars,
which are able to enrich Al while destroying O, are more
common polluters in metal-poor clusters than in metalrich ones.
5. A correlation between Al and Si in M15 indicates
particularly high-temperature hydrogen burning in the
stars that contributed the abundance pattern of the second
generation stars. This is consistent with Si enhancement
in GC stars as discussed in the literature, where it appears
to be confined to low-metallicity and high-mass clusters.
6. Besides accessing northern GCs, APOGEE is also unique
among other spectroscopic surveys because it is capable
REFERENCES
Ahn, C. P., Alexandroff, R., Allende Prieto, C., et al. 2014, ApJS, 211, 17
Alam, S., Albareti, F. D., Allende Prieto, C., et al. 2015, arXiv:1501.00963
Anders, E., & Grevesse, N. 1989, GeCoA, 53, 197
23
The Astronomical Journal, 149:153 (24pp), 2015 May
Mészáros et al.
Angelou, G. C., Stancliffe, R. J., Church, R. P., Lattanzio, J. C., & Smith, G. H.
2012, ApJ, 749, 128
Asplund, M., Grevesse, N., & Sauval, A. J. 2005, ASPC, 336, 25
Bastian, N., Lamers, H. J. G. L. M., de Mink, S. E., et al. 2013, MNRAS,
436, 2398
Bergemann, M., & Nordlander, T. 2014, arXiv:1403.3088B
Bertelli, G., Girardi, L., Marigo, P., & Nasi, E. 2008, A&A, 484, 815
Bertelli, G., Nasi, E., Girardi, L., & Marigo, P. 2009, A&A, 508, 355
Borucki, W. J., Koch, D., Basri, G., et al. 2010, Sci, 327, 977
Bovy, J., Hogg, D. W., & Roweis, S. T. 2011, AnApS, 5, 1657
Briley, M. M., Smith, V. V., King, J., & Lambert, D. L. 1997, AJ, 113, 306
Briley, M. M., Smith, V. V., Suntzeff, N. B., et al. 1996, Natur, 383, 604
Cassisi, S., Salaris, M., Pietrinferni, A., et al. 2008, ApJL, 672, L115
Carretta, E., Bragaglia, A., Gratton, R., D’Orazi, V., & Lucatello, S. 2009c,
A&A, 508, 695
Carretta, E., Bragaglia, A., Gratton, R., & Lucatello, S. 2009a, A&A, 505, 139
Carretta, E., Bragaglia, A., Gratton, R. G., et al. 2009b, A&A, 505, 117
Carretta, E., Gratton, R. G., Lucatello, S., Bragaglia, A., & Bonifacio, P. 2005,
A&A, 433, 597
Cavallo, R. M., & Nagar, N. M. 2000, AJ, 120, 1364
Charbonnel, C., & Prantzos, N. 2006, arXiv:astro-ph/0606220
Cohen, J. G., Briley, M. M., & Stetson, P. B. 2002, AJ, 123, 2525
Cohen, J. G., & Meléndez, J. 2005, AJ, 129, 303
Decressin, T., Meynet, G., Charbonnel, C., Prantzos, N., & Ekström, S. 2007,
A&A, 464, 1029
DAntona, F., Bellazzini, M., Caloi, V., et al. 2005, ApJ, 631, 868
DAntona, F., & Ventura, P. 2008, Msngr, 134, 18
D’Ercole, A., Vesperini, E., D’Antona, F., McMillan, S. L. W., & Recchi, S.
2008, MNRAS, 391, 825
Eisenstein, D. J., Weinberg, D. H., Agol, E., et al. 2011, AJ, 142, 72
Epstein, C. R., Elsworth, Y. P., Johnson, J. A., et al. 2014, ApJL, 785, L28
Freeman, K. C. 2012, in ASP Conf. Ser. 458, Galactic Archaeology: NearField Cosmology and the Formation of the Milky Way, ed. W. Aoki et al.
(San Francisco, CA: ASP), 393
García-Hernández, D. A., García-Lario, P., Plez, B., et al. 2006, Sci, 314, 1751
García-Hernández, D. A., Manchado, A., Lambert, D. L., et al. 2009, ApJ,
705, L31
Gilmore, G., Randich, S., Asplund, M., et al. 2012, Msngr, 147, 25
González Hernández, J. I., & Bonifacio, P. 2009, A&A, 497, 497
Gratton, R., Sneden, C., & Carretta, E. 2004, ARA&A, 42, 385
Gratton, R. G., Bonifacio, P., Bragaglia, A., et al. 2001, A&A, 369, 87
Gratton, R. G., Carretta, E., & Bragaglia, A. 2012, A&ARv, 20, 50
Gratton, R. G., Johnson, C. I., Lucatello, S., et al. 2011, A&A, 534, A72
Gunn, J. E., Siegmund, W. A., Mannery, E. J., et al. 2006, AJ, 131, 2332
Holtzman, J. A., Shetrone, M., Johnson, J. A., et al. 2015, arXiv:1501.04110
Ivans, I. I., Kraft, R. P., Sneden, C. S., et al. 2001, AJ, 122, 1438
Harris, W. E. 1996, AJ, 112, 1487
Ivans, I. I., Sneden, C., Kraft, R. P., et al. 1999, AJ, 118, 1273
Johnson, C. I., Kraft, R. P., Pilachowski, C. A., et al. 2005, PASP, 117, 1308
Johnson, C. I., & Pilachowski, C. A. 2012, ApJL, 754, L38
Karakas, A. I., García-Hernández, D. A., & Lugaro, M. 2012, ApJ, 8, 751
Karakas, A. I., & Lattanzio, J. C. 2003, PASA, 20, 393
Koch, A., & McWilliam, A. 2010, AJ, 139, 2289
Kraft, R. P. 1994, PASP, 106, 553
Kraft, R. P., & Ivans, I. I. 2003, PASP, 115, 143
Kraft, R. P., Sneden, C., Langer, G. E., & Prosser, C. F. 1992, AJ, 104, 645
Kurucz, R. L. 1979, ApJS, 40, 1
Lai, D. K., Smith, G. H., Bolte, M., et al. 2011, AJ, 141, 62
Lardo, C., Pancino, E., Mucciarelli, A., & Milone, A. P. 2012, A&A,
548, A107
Lee, J.-W., Carney, B. W., & Balachandran, S. C. 2004, AJ, 128, 2388
Maccarone, T. J., & Zurek, D. R. 2012, MNRAS, 423, 2
Marino, A. F., Milone, A. P., & Lind, K. 2013, ApJ, 768, 27
Martell, S. L., Smith, G. H., & Briley, M. M. 2008, AJ, 136, 2522
Meléndez, J., & Cohen, J. G. 2009, ApJ, 699, 2017
Mészáros, Sz., Allende Prieto, C., Edvardsson, B., et al. 2012, AJ, 144, 120
Mészáros, Sz., Holtzman, J., García Pérez, A. E., et al. 2013, AJ, 146, 133
Milone, A. P., Bedin, L. R., Piotto, G., et al. 2008, ApJ, 673, 241
de Mink, S. E., Pols, O. R., Langer, N., & Izzard, R. G. 2009, A&A, 507, L1
Minniti, D., Peterson, R. C., Geisler, D., & Claria, J. J. 1996, ApJ, 470, 953
O’Connell, J. E., Johnson, C. I., Pilachowski, C. A., & Burks, G. 2011, PASP,
123, 1139
Otsuki, K., Honda, S., Aoki, W., Kajino, T., & Mathews, G. J. 2006, ApJL,
641, L117
Pinsonneault, M. H., Elsworth, Y., Epstein, C., et al. 2014, arXiv:1410.2503
Piotto, G., Bedin, L. R., Anderson, J., et al. 2007, ApJL, 661, L53
Ramírez, S. V., & Cohen, J. G. 2002, AJ, 123, 3277
Pawa, S. V., & Cohen, J. G. 2003, AJ, 125, 224
Roederer, I. U., & Sneden, C. 2011, AJ, 142, 22
Shetrone, M. D. 1996, AJ, 112, 1517
Shetrone, M., Martell, S. L., Wilkerson, R., et al. 2010, AJ, 140, 1119
Smith, G. H., Shetrone, M. D., Bell, R. A., Churchill, C. W., & Briley, M. M.
1996, AJ, 112, 1511
Smith, G. H., Shetrone, M. D., & Strader, J. 2007, PASP, 119, 722
Smith, V. V., Cunha, K., Shetrone, M. D., et al. 2013, ApJ, 765, 16
Sneden, C. 1973, ApJ, 184, 839
Sneden, C., Kraft, R. P., Guhathakurta, P., Peterson, R. C., & Fulbright, J. P.
2004, AJ, 127, 2162
Sneden, C., Kraft, R. P., Prosser, C. F., & Langer, G. E. 1991, AJ, 102, 2001
Sneden, C., Kraft, R. P., Prosser, C. F., & Langer, G. E. 1992, AJ, 104, 2121
Sneden, C., Kraft, R. P., Shetrone, M. D., et al. 1997, AJ, 114, 1964
Sneden, C., Pilachowski, C. A., & Kraft, R. P. 2000, AJ, 120, 1351
Sobeck, J. S., Kraft, R. P., Sneden, C., et al. 2011, AJ, 141, 175
Steinhaus, H. 1956, Bull. Acad. Polon. Sci., 4, 801
Steinmetz, M., Zwitter, T., Siebert, A., et al. 2012, AJ, 132, 1645
Strutskie, M. F., et al. 2006, AJ, 131, 1163
Sweigart, A. V., & Mengel, J. G. 1979, ApJ, 229, 624
Wilson, J., Hearty, F., Skrutskie, M. F., et al. 2012, SPIE, 8446, 84460H
Ventura, P., D’Antona, F., Mazzitelli, I., & Gratton, R. 2001, ApJL,
550, L65
Ventura, P., di Criscienzo, M., Carini, R., & D’Antona, F. 2013, MNRAS,
431, 3642
Villanova, S., Geisler, D., & Piotto, G. 2010, ApJL, 722, L18
Yong, D., Aoki, W., & Lambert, D. L. 2006a, ApJ, 638, 1018
Yong, D., Aoki, W., & Lambert, D. L. 2006b, ApJ, 639, 918
Yong, D., Grundahl, F., D’Antona, F., et al. 2009, ApJL, 685, L62
Yong, D., Grundahl, F., Nissen, P. E., Jensen, H. R., & Lambert, D. L. 2005,
A&A, 438, 875
Yong, D., Karakas, A. I., Lambert, D. L., Chieffi, A., & Limongi, M. 2008,
ApJ, 689, 1031
Yong, D., Roederer, I. U., Grundahl, F., et al. 2014, MNRAS, 441, 3396
Zasowski, G., Johnson, J. A., Frinchaboy, P. M., et al. 2013, AJ, 146, 81
24